Sequential Benchmark

Row

About the sequential Benchmark

The sequential benchmark was done in order to compare the response time of REST and GraphQL. The test was done on two types of sample.

The few sample contains :
- 10 users - 84 products - 79 orders

The lotof sample contains :
- 100’000 users - 84 products - 300204 orders

We simulate a user doing n requests where n = {1,10,100,1000,10000} on three functions:
- getUsers (only user-microservice is involved).
- createProduct (only product-microservice is involved).
- userThatBoughtProduct (user, product and order microservices are involved).

Row

Result sample size per category, function and type

category	func	type	1	10	100	1000	10000
few	createProduct	GraphQL	50	50	50	50	50
few	createProduct	REST	50	50	50	50	50
few	getUsers	GraphQL	50	50	50	50	50
few	getUsers	REST	50	50	50	50	50
few	userThatBoughtProduct	GraphQL	50	50	50	50	0
few	userThatBoughtProduct	REST	50	50	50	50	0
lot	createProduct	GraphQL	50	50	50	50	50
lot	createProduct	REST	50	50	50	50	50
lot	getUsers	GraphQL	5	5	5	5	0
lot	getUsers	REST	5	5	5	5	0
lot	userThatBoughtProduct	GraphQL	5	5	0	0	0
lot	userThatBoughtProduct	REST	5	5	0	0	0

Response time for getUsers and createProduct for each sample size

Row

Response time for getUsers and createProduct for each sample size (boxplot)

Concurrent benchmark (autocannon)

Row

success/failure distribution

Response time per concurrent user

Concurrent benchmark (JMeter)

A concurrency benchmark was done with JMeter to test server performance and response time. The sample used is the few sample one. The test has not been done with the lotof sample because response times are too high.

Row

Distribution of success and failure per function

func	type	success	failure	success percent (%)
addProduct	graphql	419973	3158109	11.74
addProduct	rest	432346	4175496	9.38
getUsers	graphql	429955	5713588	7.00
getUsers	rest	429939	5150337	7.70
productBoughtByUser	graphql	140286	12510546	1.11
productBoughtByUser	rest	178045	7617850	2.28